Effective hyperspectral image classification based on segmented PCA and 3D-2D CNN leveraging multibranch feature fusion

Afjal, Masud Ibn; Mondal, Md. Nazrul Islam; Mamun, Md. Al

doi:10.1080/14498596.2024.2305119

Cited by 4 publications

(6 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Spatial features include textures, edges, key points, shapes, etc., which can be extracted by traditional feature extractors [24,28,32,34,37] and deep neural networks [68,73,75,76,86], from each band of an HSI cube. But the discrimination of these features may be weak because relevant spatial areas tend to be ignored.…”

Section: Extraction Of Discriminating Spatial Featuresmentioning

confidence: 99%

“…Due to the reflective discrepancy, the spatial structures in each band may be different. To improve the efficiency of feature extraction, PCA was utilized to extract the prime spatial information [40,41,86]. But this measure cannot take spectral correlation into account due to the loss of band order.…”

Section: Common Spatial Features Traditional Feature Descriptorsmentioning

confidence: 99%

“…To extract abstract and deep spectral features, many algorithms based on deep learning have been proposed in recent years. CNNs [54], as one of the most popular models, can extract various local spatial features with different convolutional kernels and retain semantic and prominent spatial information with pooling operations [72][73][74][75][76][77][78][79][80][81][82][83][84][85][86]. Many previous models attempted integration with CNNs, such as ConvAEs [70,71,87], Con-vRNNs [89], ConvLSTM networks [90], GCNs [91], and DCGANs [106][107][108], to supplement deep spatial feature representation for classification.…”

Section: Deep Network Architecturesmentioning

confidence: 99%

“…The first one arranges the spectral subnetwork and spatial subnetwork sequentially [76,86,88,95,97,209]. This kind of model usually collects global spectral information with 1D convolutional layers or RNNs and then exploits 2D convolutions to process spatial features for classification.…”

Section: Sequential Spectral-spatial (Seq-ea) Modelsmentioning

confidence: 99%

“…Different types of features [75,77,78] and probability scores [79,80] are fused for better classification results. Benefitting from cubic kernels, three-dimensional (3D) CNNs can directly extract the representative spectral-spatial features from HSI cubes and fuse them in each stage of the model [72,[81][82][83][84][85][86]. Therefore, CNNs have been introduced to previous models to construct convolutional auto-encoders (ConvAEs) [87,88], convolutional recurrent neural networks (ConvRNNs) [89], convolutional long short-term memory (ConvLSTM) networks [90], graph convolutional networks (GCNs) [91], etc.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Discriminating Spectral–Spatial Feature Extraction for Hyperspectral Image Classification: A Review

Li,

Wang,

Cheikh

2024

Sensors

View full text Add to dashboard Cite

Hyperspectral images (HSIs) contain subtle spectral details and rich spatial contextures of land cover that benefit from developments in spectral imaging and space technology. The classification of HSIs, which aims to allocate an optimal label for each pixel, has broad prospects in the field of remote sensing. However, due to the redundancy between bands and complex spatial structures, the effectiveness of the shallow spectral–spatial features extracted by traditional machine-learning-based methods tends to be unsatisfying. Over recent decades, various methods based on deep learning in the field of computer vision have been proposed to allow for the discrimination of spectral–spatial representations for classification. In this article, the crucial factors to discriminate spectral–spatial features are systematically summarized from the perspectives of feature extraction and feature optimization. For feature extraction, techniques to ensure the discrimination of spectral features, spatial features, and spectral–spatial features are illustrated based on the characteristics of hyperspectral data and the architecture of models. For feature optimization, techniques to adjust the feature distances between classes in the classification space are introduced in detail. Finally, the characteristics and limitations of these techniques and future challenges in facilitating the discrimination of features for HSI classification are also discussed further.

show abstract